[Dnsmasq-discuss] dnsmasq exists with sigterm following bootp messages

[Thomas Ballantyne]

Hi Simon,

Seeing an issue where BOOTP messages are crashing dnsmasq. dnsmasq exits
with a SIGTERM.

I tried adding

dhcp-ignore=bootp


to dnsmasq.more.conf but that did not solve the problem.

I've tested on Ubuntu 14.0.4, with dnsmasq 2.75.


We run dnsmasq within a virtual machine image of Ubuntu 14. We connect
devices under test (set top boxes, smart TVs, consoles, etc.) to the VM via
a USB dongle. The device under test attempts to lease an IP address from
dnsmasq on the host machine (192.168.144.xx) and then connect to the
Netfllix cloud (10.10.xx.xx.) Our Network Agent software running on the
host acts like a simple two port router, reconfigures the device under
test, and does additional dns overrides to allow connections to our test
and production infrastructure.

We see this issue specifically when connecting a partner Smart TV to the
host running dnsmasq via USB dongle. The partner TV runs Linux/Android
OS. Other devices under the same configuration do not show this behavior.
We've reported the issue to the partner as well. They indicate they do not
see the issue when bypassing dnsmasq. Problem happens anywhere from 30-50%
of hard reboots. The end result is the TV fails to lease an IP address.
I've also seen the USB dongle issue BOOTP messages causing dnsmasq to exit
with sigterm.

'service restart dnsmasq' works to bring dnsmasq back after the failure.

I've attached a portion of the dnsmasq logs (30 lines before sigterm.)

Anything we can do here? Let me know if I can provide more detail.

Thanks,
Thomas


Thomas Ballantyne
Operations Engineer, Netflix

[Thomas Ballantyne]

Hi,

Any feedback here?

Thanks,
Thomas

Post by Thomas Ballantyne
Hi Simon,
Seeing an issue where BOOTP messages are crashing dnsmasq. dnsmasq exits
with a SIGTERM.
I tried adding
dhcp-ignore=bootp
to dnsmasq.more.conf but that did not solve the problem.
I've tested on Ubuntu 14.0.4, with dnsmasq 2.75.
We run dnsmasq within a virtual machine image of Ubuntu 14. We connect
devices under test (set top boxes, smart TVs, consoles, etc.) to the VM via
a USB dongle. The device under test attempts to lease an IP address from
dnsmasq on the host machine (192.168.144.xx) and then connect to the
Netfllix cloud (10.10.xx.xx.) Our Network Agent software running on the
host acts like a simple two port router, reconfigures the device under
test, and does additional dns overrides to allow connections to our test
and production infrastructure.
We see this issue specifically when connecting a partner Smart TV to the
host running dnsmasq via USB dongle. The partner TV runs Linux/Android
OS. Other devices under the same configuration do not show this behavior.
We've reported the issue to the partner as well. They indicate they do not
see the issue when bypassing dnsmasq. Problem happens anywhere from 30-50%
of hard reboots. The end result is the TV fails to lease an IP address.
I've also seen the USB dongle issue BOOTP messages causing dnsmasq to exit
with sigterm.
'service restart dnsmasq' works to bring dnsmasq back after the failure.
I've attached a portion of the dnsmasq logs (30 lines before sigterm.)
Anything we can do here? Let me know if I can provide more detail.
Thanks,
Thomas
Thomas Ballantyne
Operations Engineer, Netflix

[Simon Kelley]

Why do you think that this is BOOTP related? I can't see any evidence
that your clients are sending bootp requests.

Background: BOOTP is an old protocol that pre-dates DHCP. DHCP packets
are actually BOOTP packets with lots of extra options. From the logs,
all the packets that dnsmasq is seeing have the extra options and are
DHCP packets.

Are you sure that dnsmasq is lying? It looks to me like something is
actually sending a SIGTERM to dnsmasq and it's doing what it supposed
to do when that happens and executing an orderly shutdown. In most,
but not all of the logs, it seems that dnsmasq is getting a SIGHUP
which causes it to re-read its configuration, and then also getting
SIGTERM which causes the shutdown immediately after.

I'd be looking very carefully at whatever bit of scripting is sending
those SIGHUPs and seeing if it can also send SIGTERM.


Cheers,

Simon.

Post by Thomas Ballantyne
Hi,
Any feedback here?
Thanks, Thomas
On Tue, Dec 1, 2015 at 12:16 PM, Thomas Ballantyne

Post by Thomas Ballantyne
Hi Simon,
Seeing an issue where BOOTP messages are crashing dnsmasq.
dnsmasq exits with a SIGTERM.
I tried adding
dhcp-ignore=bootp
to dnsmasq.more.conf but that did not solve the problem.
I've tested on Ubuntu 14.0.4, with dnsmasq 2.75.
We run dnsmasq within a virtual machine image of Ubuntu 14. We
connect devices under test (set top boxes, smart TVs, consoles,
etc.) to the VM via a USB dongle. The device under test attempts
to lease an IP address from dnsmasq on the host machine
(192.168.144.xx) and then connect to the Netfllix cloud
(10.10.xx.xx.) Our Network Agent software running on the host
acts like a simple two port router, reconfigures the device
under test, and does additional dns overrides to allow
connections to our test and production infrastructure.
We see this issue specifically when connecting a partner Smart TV
to the host running dnsmasq via USB dongle. The partner TV runs
Linux/Android OS. Other devices under the same configuration do
not show this behavior. We've reported the issue to the partner
as well. They indicate they do not see the issue when bypassing
dnsmasq. Problem happens anywhere from 30-50% of hard reboots.
The end result is the TV fails to lease an IP address. I've also
seen the USB dongle issue BOOTP messages causing dnsmasq to exit
with sigterm.
'service restart dnsmasq' works to bring dnsmasq back after the failure.
I've attached a portion of the dnsmasq logs (30 lines before
sigterm.)
Anything we can do here? Let me know if I can provide more
detail.
Thanks, Thomas
Thomas Ballantyne Operations Engineer, Netflix

_______________________________________________ Dnsmasq-discuss
http://lists.thekelleys.org.uk/mailman/listinfo/dnsmasq-discuss